Dynamic supporting type self-adaptive scooter based on artificial intelligence

Abstract

The invention belongs to the technical field of intelligent scooters, provides a dynamic supporting type self-adaptive scooter based on artificial intelligence, and innovatively solves the defects of an existing scooter in the conversion process of a sitting posture and a standing posture by combining an intelligent AI model, a 360-degree somatosensory camera, sensor data and an electroactive polymer supporting system. Through the self-adaptive adjusting function of the AI model, the system can accurately carry out dynamic supporting adjustment in real time according to the posture, pressure and body shape of the user, and comfort, stability and health of each stage are ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent mobility scooters, and particularly relates to a dynamic support type adaptive mobility scooter based on artificial intelligence. Background Art

[0002] With the increase in the aging population and the number of lower limb disabled patients, as an important auxiliary device, mobility scooters have been widely used in the daily lives of the elderly, disabled people, and those with limited mobility. The existing types of mobility scooters include traditional wheelchairs, electric wheelchairs, and electric mobility scooters, etc., which mainly provide basic mobility functions through mechanical structures or electric drives. However, these existing mobility scooters still have the following several problems and disadvantages in actual use:

[0003] 1. Difficulty in the transition between sitting and standing postures

[0004] Traditional mobility scooters usually only support one mode of sitting or standing, or provide the function of sitting and standing conversion, but the conversion process often requires external assistance or strenuous operation. For lower limb disabled people, the conversion from sitting to standing is not only difficult, but also may cause additional pressure on the spine, knee joints, and other key parts without proper support, increasing the body burden and even causing related diseases (such as low back pain, arthritis, etc.).

[0005] 2. The support system is not flexible enough

[0006] Most of the existing mobility scooters adopt rigid support devices, such as fixed backrests, seat cushions, and armrests. Although these rigid support devices can provide a certain degree of stability, they cannot be dynamically adjusted according to the user's posture changes, resulting in discomfort for users during long-term use, especially during sitting and standing conversion, long-term sitting or standing postures, where the pressure on some parts of the body is too large, easily causing discomfort or health problems (such as pressure sores, joint pain, etc.).

[0007] 3. Insufficient intelligence

[0008] The intelligence level in existing mobility scooters is generally low, and the vast majority rely on manual operation by users to complete operations such as mode conversion and support adjustment. Although some high-end mobility scooters have tried to introduce electric drive and automatic adjustment functions, most of these systems lack a real-time perception and intelligent response mechanism for the user's posture changes and cannot automatically adjust the support structure according to the user's real-time state, resulting in insufficient guarantee of the comfort and safety of users during actual use.

[0009] 4. Insufficient comfort and ergonomic design

[0010] Most existing mobility scooter designs only consider the mobility function and lack in-depth consideration of ergonomics. Especially for users who use mobility scooters for a long time, components such as the seat, backrest, and armrests of traditional mobility scooters may not adapt to the body shapes and postural needs of different users, resulting in discomfort or health problems during long-term use. Summary of the Invention

[0011] To solve the above technical problems, the present invention provides a dynamically supported adaptive mobility scooter based on artificial intelligence to solve the problems in the prior art. The technical solution adopted by the present invention is as follows:

[0012] A dynamically supported adaptive mobility scooter based on artificial intelligence, comprising a vehicle body, a 360° body-sensing camera system, an intelligent AI model, and an electroactive polymer support system;

[0013] The 360° body-sensing camera system is arranged on the vehicle body and is used to capture the user's posture;

[0014] The electroactive polymer support system is arranged on the vehicle body;

[0015] The intelligent AI model is connected to the 360° body-sensing camera system and the electroactive polymer support system; the intelligent AI model is used to adjust the electroactive polymer support system according to the user's posture to adapt to different postures of the user.

[0016] Further, the 360° body-sensing camera system includes two cameras, which are respectively arranged at the upper and lower positions of the vehicle body and are used to capture the upper body posture and lower body posture of the user.

[0017] Further, the vehicle body includes a seat, a front baffle, and a footboard; the front end of the footboard is fixedly connected to the bottom of the front baffle, and the rear end of the footboard is fixedly connected to the seat; a rear wheel is arranged below the seat, and a front wheel is arranged below the footboard.

[0018] Further, the electroactive polymer support system includes a chest support, leg pads, foot pads, and a rear back pad installed on the vehicle body; the chest support and the rear back pad are arranged opposite to each other and are spaced apart.

[0019] Further, the intelligent AI model includes a vision transformer, a graph neural network, and a long short-term memory network.

[0020] Further, the input of the vision transformer is the user image data of the 360° body-sensing camera system, which is used to extract global features from the image and capture the posture changes of the user;

[0021] The visual transformer divides each frame of image into image patches of a fixed size through image chunking. Each image patch is converted into a vector through linear projection and position encoding is added to preserve the spatial position information of the image patches.

[0022] The visual transformer uses the self-attention mechanism for multi-layer feature learning to capture the global semantic information in the image.

[0023] The visual transformer outputs high-level features of the image extracted by multiple self-attention layers.

[0024] Further, the input of the graph neural network is the user image data of the 360° somatosensory camera system, which is used to capture the spatial relationship and dependence between human joints.

[0025] The graph neural network initializes each human joint as a node in the graph through node initialization. The features of the node include the coordinates, angles, and speed information of the joint.

[0026] The graph neural network propagates information through the graph convolutional layer to calculate the spatial dependence relationship between each joint.

[0027] The graph neural network generates the spatial dependence features between joints and transmits them to the long short-term memory network for temporal modeling.

[0028] Further, the input of the long short-term memory network comes from the image features and spatio-temporal features of the visual transformer and the graph neural network; it is used to model the movement and posture changes of the human body in the time dimension and capture long-term dependencies.

[0029] The long short-term memory network receives the spatio-temporal features from the visual transformer and the graph neural network and processes the time series data.

[0030] The long short-term memory network captures the time series dependence through the internal gating mechanism.

[0031] The long short-term memory network generates the posture category and support strength and feeds them back to the electroactive polymer support system to adjust the electroactive polymer support system.

[0032] The present invention has the following beneficial effects:

[0033] (1) Automatic sitting-standing conversion and support adjustment: Existing mobility scooters usually rely on manual operation by the user or external assistance for sitting-standing conversion, which is a cumbersome process and may pose certain risks. The present invention integrates a 360° somatosensory camera and intelligent AI algorithms, which can automatically analyze and predict the user's posture changes and adjust the support structure in real time, making the sitting-standing conversion smoother and more convenient, greatly improving the user's independence and comfort. Without external help, the user can transition from a sitting position to a standing position more easily.

[0034] (2) Personalized and dynamic support system: This invention uses electroactive polymers (EAP) as flexible support materials. Compared with traditional fixed support systems, EAP can automatically adjust the support strength according to the user's real-time posture changes, providing personalized support. This dynamic adjustment ability enables the mobility scooter to adapt to the needs of different users, ensuring that the user can obtain the best comfortable support whether in a sitting position, a standing position or during the transition process. Especially during the sitting-standing conversion process, the EAP material can effectively reduce the body's pressure points and avoid compression and discomfort in areas such as the back and knees.

[0035] (3) Intelligent AI model and real-time feedback: Through the collaborative work of sensor fusion technology and the AI model, the system can capture the user's posture and joint pressure in real time and make support adjustments in a timely manner. The intelligence level of traditional mobility scooters is relatively low, usually relying only on manual adjustment of the support strength. However, the AI model of this invention can not only automatically sense the user's posture changes, but also adjust the support structure in real time according to multi-modal data to ensure that the user gets the best support in different postures. In addition, the AI model can continuously learn the user's posture characteristics to achieve personalized customization and further improve comfort.

[0036] (4) Omnidirectional posture monitoring and optimization: This invention uses a 360° body sensor camera to monitor the upper and lower body postures of the user in real time, ensuring that the user's motion data can be captured without dead angles. This enables the system to more accurately evaluate the user's sitting and standing postures and the needs during the posture transition process, optimize the response of the support system, and avoid the traditional mobility scooter from only obtaining single-angle data through simple sensors, thereby enhancing the intelligence level of the system.

[0037] (5) Health protection and comfort improvement: Through the real-time analysis of posture changes, this invention can reduce the pressure on areas such as the back and joints that may be caused by long-term sitting and standing, preventing health problems such as bedsores and joint injuries. By adjusting the strength and form of the support system, the system can adapt to the user's body type and health condition, thereby significantly improving comfort and reducing discomfort. Compared with the fixed support device of traditional mobility scooters, the intelligent support system of this invention conforms more to the ergonomic principle and is more beneficial to the user's health.

[0038] (6) Improving user independence and quality of life: This invention enables users to complete the sitting-standing conversion more independently, no longer requiring the assistance of external personnel, greatly enhancing the independence of people with limited mobility. The intelligent posture adjustment and dynamic support system further improve the user's comfort and improve the quality of life of those who have long relied on traditional mobility scooters. For the lower limb disabled and the elderly population, especially those with long-term sitting and lying needs, this invention can provide a more comfortable, convenient and safe mobility solution.

[0039] (7) Promote the intelligent development of auxiliary devices: The present invention promotes the development of intelligent auxiliary devices by integrating a variety of advanced technologies (such as artificial intelligence, image processing, flexible materials, etc.). In addition to its application in the field of mobility scooters, the technical solution of the present invention can also be extended to other medical auxiliary devices (such as intelligent wheelchairs, rehabilitation equipment, etc.), with high technical universality and market application potential.

[0040] (8) By combining technologies such as 360° somatosensory cameras, intelligent AI algorithms, and electroactive polymers (EAPs), the present invention innovatively solves the deficiencies of existing mobility scooters in sitting-standing conversion and support adjustment, achieving fully automatic, personalized, and dynamic support adjustment. Compared with the prior art, the present invention has the advantages of being more intelligent, comfortable, healthy, and flexible, providing users with a better mobility experience, improving the quality of life, and promoting the development of intelligent auxiliary device technology. Brief Description of the Drawings

[0041] Figure 1 is a front-side schematic diagram of the overall structure of the present invention;

[0042] Figure 2 is a rear-side schematic diagram of the overall structure of the present invention;

[0043] Figure 3 is a diagram of the parameters and evaluation results of the sub-model;

[0044] Figure 4 is the support adjustment and AI model workflow in the sitting position;

[0045] Figure 5 is the support adjustment and AI model workflow in the standing position;

[0046] Figure 6 is the support adjustment and AI model during the sitting-standing conversion process;

[0047] Figure 7 is a diagram of the AI model adaptively adjusting parameters;

[0048] Figure 8 is the self-adaptive workflow diagram of the AI model;

[0049] Figure 9 is a schematic diagram of the hardware layout of the present invention. Detailed Embodiments

[0050] The following will combine the Figures 1-9 in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. If not specifically specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0051] The objective of the present invention is to provide an AI-based dynamic support type adaptive mobility scooter, aiming to solve the inconvenience during the sitting and standing conversion process of existing mobility scooters and the deficiencies of the support system. The specific objectives are as follows:

[0052] Realize the adaptive sitting and standing conversion function: During the sitting and standing conversion process of existing mobility scooters, manual operation or external assistance is often relied upon. Through the introduction of a 360° body-sensing camera and intelligent AI algorithms, the present invention can automatically track the user's posture changes, analyze in real time, and precisely adjust the support system, helping the user seamlessly transition from a sitting position to a standing position and providing a more convenient sitting and standing conversion experience.

[0053] Provide a dynamic and intelligent support system: The present invention combines electroactive polymers (EAP) and intelligent algorithms, which can automatically adjust the support structure according to the user's real-time posture changes, thereby ensuring that key parts such as the waist, knees, and neck obtain the best support and avoiding discomfort or health problems caused by long-term sitting or standing, such as back pain and pressure sores.

[0054] Improve the user's comfort and independence: By introducing an intelligent AI model, the present invention can process data from the 360° camera and other sensors in real time, dynamically adjust the support intensity to adapt to different body shapes, postures, and needs, greatly improving the user's comfort and independence, reducing the dependence on others' help, and enhancing the quality of life.

[0055] Promote the user's health: The present invention aims to reduce the pressure on the body caused by long-term sitting or standing postures through accurate posture recognition and support adjustment, help the user maintain a healthy posture, and prevent problems such as spinal deformity and joint injury, thereby improving the overall health level of the user.

[0056] Promote the development of intelligent assistive device technology: The present invention not only provides innovative solutions in the field of mobility scooters but also provides a reference for other intelligent assistive devices (such as wheelchairs and rehabilitation devices), promoting the application and development of artificial intelligence, sensor technology, and new flexible materials (such as EAP) in health management and assistive devices.

[0057] Such as Figure 1 、 Figure 2 , an AI-based dynamic support type adaptive mobility scooter, comprising a vehicle body, a 360° body-sensing camera system, an intelligent AI model, and an electroactive polymer support system;

[0058] The 360° body-sensing camera system is arranged on the vehicle body for capturing the user's posture;

[0059] The electroactive polymer support system is arranged on the vehicle body;

[0060] The intelligent AI model is connected to the 360° somatosensory camera system and the electroactive polymer support system; the intelligent AI model is used to adjust the electroactive polymer support system according to the user's posture to adapt to different postures of the user.

[0061] Specifically, the specific technical content of each part is as follows:

[0062] 360° somatosensory camera system, which includes Camera 1: Camera 1 is installed in the center of the armrest bracket to capture the posture data of the user's upper body (shoulders, arms, neck, etc.). Especially during the sitting-standing conversion, it provides accurate upper body movement data; Camera 2: Camera 2 is installed under the seat and can capture the lower body posture (legs, knees, ankles, etc.) from below to ensure full-range tracking of the posture, especially during the conversion between sitting and standing postures.

[0063] Intelligent AI model, the AI model combines 360° somatosensory camera and other sensor data, uses machine learning algorithms to analyze the user's posture in real time, predicts the joint pressure and dynamic changes during the sitting-standing conversion, and automatically adjusts the support structure.

[0064] Core functions of AI:

[0065] Posture detection and prediction: The AI analyzes and predicts the sitting-standing conversion process based on the user's posture data captured by the camera, helps to adjust the support structure in a timely manner, and ensures that the ergonomic principles are followed.

[0066] Real-time support adjustment: Through intelligent algorithms, the AI can adjust the deformation of electroactive polymer (EAP) materials in real time according to the pressure on different joints (such as the waist, knees, ankles, etc.), so as to provide dynamic support.

[0067] Electroactive polymer (EAP) support system, the EAP material can generate deformation under the action of an electric field and is suitable for human support parts (such as the waist, knees, neck, etc.). Dynamic support: By adjusting the electric field strength, the EAP material automatically deforms to provide personalized support. Posture adjustment: According to the real-time analysis of the AI model, the EAP material automatically adjusts the support strength to help the user obtain the best support during the sitting-standing conversion. Lightweight and flexible: The EAP material has good adaptability and can fit the natural curve of the human body, avoiding the discomfort caused by fixed support devices.

[0068] The present invention also includes a sensor system. By combining multi-modal data from a 360° camera, seat pressure sensors, and motion sensors, the AI can accurately perceive the user's posture changes in real time and automatically adjust the support system.

[0069] One of the core innovations of the present invention is the intelligent AI model, which combines a Vision Transformer (ViT), a Graph Neural Network (GNN), and a Long Short-Term Memory Network (LSTM), and automatically adjusts a flexible support system (such as electroactive polymers, EAP) by real-time analyzing user posture data. The following are the detailed AI model structure, architecture, and working principle:

[0070] AI model architecture: The AI model of the present invention consists of three main parts: visual feature extraction, posture and spatial relationship modeling, and temporal modeling and support adjustment. These three parts use a Vision Transformer (ViT), a Graph Neural Network (GNN), and a Long Short-Term Memory Network (LSTM) respectively to process different data types and dimensions.

[0071] Vision Transformer (ViT), tasks of the Vision Transformer: Feature extraction of image data; Input of the Vision Transformer: User image data from a 360° body sensor camera; Function of the Vision Transformer: Extract global features from images and capture user posture changes. Working principle of the Vision Transformer: Image chunking: Each frame of the image is segmented into fixed-size image patches, and each patch of the image is converted into a vector through linear projection. Position encoding: Add position encoding to retain the spatial position information of the image patches. The Vision Transformer also includes a self-attention mechanism: Use the self-attention mechanism to perform multi-layer feature learning and capture the global semantic information in the image (such as the movement of the upper body, the positions of the arms and shoulders). Output of the Vision Transformer: Extract high-level features of the image through multiple self-attention layers for use by subsequent processing modules.

[0072] Graph Neural Network (GNN), tasks of the Graph Neural Network: Modeling the spatial dependency relationships of posture data; Input of the Graph Neural Network: Posture data from the camera (such as joint coordinates); Function of the Graph Neural Network: Capture the spatial relationships and dependencies between human joints. Working principle of the Graph Neural Network: The Graph Neural Network includes node initialization: Each human joint (for example, the neck, shoulders, elbows, etc.) is used as a node in the graph, and the features of the node include information such as the coordinates, angles, and speeds of the joints. The Graph Neural Network includes a Graph Convolutional Layer (GCN): Propagate information through the Graph Convolutional Layer to calculate the spatial dependency relationships between each joint. The Graph Convolutional Layer aggregates the information of neighbor nodes (i.e., adjacent joints) to capture the movement interaction relationships between joints. Output of the Graph Neural Network: Generate the spatial dependency features between joints and further transmit them to the LSTM module for temporal modeling.

[0073] Long Short-Term Memory Network (LSTM), tasks of the LSTM: modeling time series data and pose prediction; inputs of the LSTM: image features and spatio-temporal features from ViT and GNN; functions of the LSTM: modeling the movement and pose changes of the human body in the time dimension and capturing long-term dependencies. Working principle of the LSTM: The LSTM includes time series input: The LSTM receives spatio-temporal features from ViT and GNN and processes time series data. The LSTM includes state update: The LSTM updates its state through internal gating mechanisms (input gate, forget gate, output gate) to capture long-time dependencies (such as the transition process from sitting to standing). Output of the LSTM: Generates prediction results such as pose categories and support strength and feeds them back to the support adjustment system.

[0074] Such as Figure 3 , the working principle of the AI model of the present invention:

[0075] The working process of the AI model starts from inputting image data and pose data, goes through multiple steps such as image feature extraction, spatial relationship modeling, time series processing, and dynamic support adjustment, and finally outputs appropriate support adjustment parameters.

[0076] Input of image data: Images from a 360° somatosensory camera are fed into a Vision Transformer (ViT) to extract global image features (such as pose data of the user's upper body, neck, and shoulders).

[0077] Input of pose data: The joint coordinate data extracted from the image is input into a Graph Neural Network (GNN) to establish spatial dependencies between human joints and capture the interactions between various body parts (such as knees, spine, ankles, etc.).

[0078] Processing of time series data: The spatio-temporal features output by ViT and GNN are fed into a Long Short-Term Memory Network (LSTM) to capture the time series information of pose changes, such as the dynamic process from sitting to standing, and predict the user's next action.

[0079] Adaptive support adjustment: Based on the output of the LSTM, the AI model adjusts the deformation strength of electro-active polymer (EAP) materials in real time to adjust the support structure. For example, if the AI detects excessive pressure on the user's lower back, the EAP material will automatically deform to provide more lumbar support and relieve the back burden.

[0080] Such as Figure 3 , the adaptive ability of the AI model of the present invention is mainly reflected in the following aspects:

[0081] Real-time Posture Analysis: Through sensor data (posture data from a 360° camera and sensors), the AI model can analyze the user's posture changes in real time. For example, when the user changes from a sitting position to a standing position, the system will automatically predict the sitting-to-standing transition process based on data such as joint angles and pressure distribution, and adjust the support system to ensure comfort.

[0082] Adaptive Support Adjustment: AI Dynamic Adjustment: The AI model adjusts the deformation strength of electro-active polymer (EAP) materials by real-time monitoring of joint pressure and posture. For example, if the system detects excessive pressure in the spinal area, the EAP material will increase its deformation to provide more support. Through electric field regulation, the deformation of EAP can be precisely controlled, not only with rapid response but also without any mechanical devices.

[0083] Learning and Optimization: Parameter Adjustment: The AI model will automatically adjust various parameters (such as posture classification thresholds, joint pressure judgment criteria, etc.) through continuous learning and optimization, enabling the system to better adapt to the needs of different users. For example, for users with a tall stature, the AI can automatically adjust the support strength to provide optimal support during different sitting-to-standing transitions.

[0084] Personalized Customization: Personalized Learning: The AI system can perform personalized customization according to each user's posture characteristics, body type, and needs. For example, for users with special physiological structures or health problems, the AI can adjust the support strength and transition speed to provide more appropriate posture support.

[0085] Through a deep learning model of Vision Transformer (ViT), Graph Neural Network (GNN), and Long Short-Term Memory Network (LSTM), the present invention provides an intelligent system capable of real-time analyzing, predicting, and adjusting the user's posture. The AI model can not only accurately capture the user's posture changes but also adaptively adjust the support strength to ensure the comfort and safety of the user during different sitting-to-standing transitions. Through sensor fusion and intelligent feedback, the present invention endows the mobility scooter with high intelligence, flexibility, and self-adaptability, greatly improving the user's independence and quality of life.

[0086] The algorithm of the present invention is as follows:

[0087] Processing Flow:

[0088] Image Blocking:

[0089]

[0090] where P = 16 is the block size and N = HW / P 2 is the number of blocks; Position Encoding:

[0091] PE (pos,2i)= sin(pos / 10000 2i / d )PE (pos,2i+1) = cos(pos / 10000 2i / d ) where d is the embedding dimension and pos is the block position index;

[0092] Multi-head self-attention:

[0093]

[0094] Introduce pose key point attention weights;

[0095]

[0096] where is the ReLU activation, σ is the Sigmoid function Graph Neural Network (GNN) algorithm:

[0097] Input: Joint coordinates J ∈ R M×3 ;

[0098] Graph construction:

[0099] Node features:

[0100]

[0101] Contain 3D coordinates, joint angles, velocities;

[0102] Edge weight calculation:

[0103]

[0104] where σ = 0.2 controls the joint connection strength;

[0105] Graph convolutional layer:

[0106]

[0107] Introduce time difference features:

[0108]

[0109] Long Short-Term Memory Network (LSTM);

[0110] Gating mechanism:

[0111] i t = σ(W xi x t + W hi h t-1 + b i )

[0112] f t = σ(Wxf x t +W hf h t-1 +b f )

[0113] o t =σ(W xo x t +W ho h t-1 +b o )

[0114]

[0115]

[0116] h t =o t ☉tanh(C t )

[0117] Multi-modal feature fusion gate:

[0118]

[0119] Multi-task learning objective function:

[0120]

[0121] Wherein:

[0122] Posture classification loss:

[0123]

[0124] Pressure prediction loss:

[0125]

[0126] Motion smoothing constraint:

[0127]

[0128] Real-time control algorithm:

[0129] EAP driving voltage mapping function:

[0130]

[0131] Wherein: s_k is the predicted pressure value of the k-th support point; the parameters α = 80V, β = 0.2, γ = 10V·s are calibrated through experiments.

[0132] The following details the specific working process of the present invention:

[0133] (1) Sitting posture state: In the sitting posture state, the center of gravity of the human body is mainly concentrated on the buttocks, lower back and thighs. Prolonged sitting is likely to cause compression on the spine, buttocks and thighs. Therefore, dynamic adjustment is required through an intelligent support system. Traditional mobility scooters usually adopt a rigid support structure and cannot adjust the support intensity in real time according to the user's posture changes, resulting in poor comfort and even possible negative impacts on physical health (such as pressure sores, back pain, etc.).

[0134] Such as Figure 4 , the AI model of the present invention predicts the user's sitting posture by analyzing the data from a 360° somatosensory camera and sensors (pressure, movement, etc.) in real time, and dynamically adjusts the support device. The working process of the AI model is as follows:

[0135] Image feature extraction (ViT): The Vision Transformer (ViT) extracts global features from the user's sitting posture image captured by the camera, such as the posture of the user's upper body (shoulders, neck) and the angles of the lower body (knees, buttocks). Through the self-attention mechanism, ViT captures the spatial relationships in the image and provides global semantic information for subsequent data analysis.

[0136] Spatial dependence modeling (GNN): GNN captures the interactions between them in the sitting posture by modeling the spatial relationships between human joints (such as knee joints, buttocks and spine). When the AI detects a large pressure on the back or thigh in the sitting posture, the system will predict which parts need additional support based on the spatial features extracted by GNN, and automatically adjust the support intensity through the support device.

[0137] Temporal modeling (LSTM): LSTM performs temporal analysis based on the features from ViT and GNN, and learns the subtle changes of the user in the sitting posture. When the AI recognizes a small adjustment of the user's sitting posture, LSTM will predict the future posture and adjust the support parameters according to the real-time data, such as the seat depth, the inflation volume of the seat cushion air cushion, etc.

[0138] Figure 4 shows how the AI extracts sitting posture features from camera and sensor data. According to the posture changes, how the EAP support system adjusts the support force to ensure uniform pressure distribution on the waist, buttocks and thighs.

[0139] The AI model realizes adaptive dynamic support adjustment in the following ways:

[0140] Automatic adjustment of lumbar support: When the AI detects that the user's waist is under excessive pressure, the EAP support material automatically deforms by adjusting the electric field strength to provide additional lumbar support and relieve spinal pressure.

[0141] Thigh and Hip Pressure Regulation: Pressure sensors continuously monitor the pressure in the hip and thigh areas. The AI model adjusts the seat depth or activates the air cushion support according to the pressure distribution to help disperse the pressure and avoid health problems caused by long-term sitting postures.

[0142] Adaptability of Flexible Support: The EAP material can change the support strength in real-time according to the AI feedback, ensuring that users maintain a natural sitting posture curve during sitting and avoiding a sense of compression.

[0143] (2) Standing State: When standing, the center of gravity of the human body is concentrated on the legs, and the knees, spine, and ankles bear relatively large pressure. Traditional mobility aids lack effective support adjustment during the standing process. Users may need to rely on external support or their own efforts to maintain balance, which may lead to fatigue or discomfort.

[0144] Such as Figure 5 , the AI model also plays a crucial role during the standing process to ensure that users remain comfortable and stable while standing. The AI monitors the user's standing posture in real-time and adaptively adjusts the support system based on the data.

[0145] Standing Posture Recognition and Prediction (ViT): The Vision Transformer (ViT) analyzes the changes in the user's upper and lower body during standing, identifies the postures of parts such as the knees and spine, and predicts possible imbalances or discomforts during standing.

[0146] Knee and Spine Support (GNN): The GNN evaluates the burden on these parts during standing by modeling the spatial relationship between the knees and the spine. When the AI detects that the pressure on the knees is too high, the GNN predicts possible posture deviations based on the data and adjusts the knee support.

[0147] Standing Stability and Prediction (LSTM): The LSTM predicts the changes in the user's standing posture based on historical standing data and adjusts the support structure in advance. For example, when the user stands for a long time, the LSTM monitors the fatigue of the knees and ankles and automatically adjusts the support strength to help reduce joint pressure.

[0148] Figure 5 Demonstrates how the AI dynamically adjusts the deformation of the EAP support material through real-time posture detection, analyzing the pressure on the knees and spine. It includes the role of the EAP support system in the standing position, how to adjust the support strength according to the knee joint angle and spine posture to ensure balance and comfort.

[0149] (3) Support adjustment during the sitting-standing conversion: During the conversion from sitting to standing, the AI system analyzes the data from the 360° somatosensory camera in real time to predict the user's movement trajectory and posture changes. When the system detects that the user is about to stand up, the AI will adjust the support system in advance (such as enhancing the support of the waist and knees) to ensure a smooth and comfortable sitting-standing conversion process.

[0150] For example Figure 6 , during the conversion from sitting to standing, the EAP material will dynamically adjust its deformation to provide additional support for the waist, knees, spine and other parts as needed. At the same time, the system will also control the adjustment of the seat and armrests so that the user can maintain the correct posture when standing.

[0151] Figure 6 Shows the support adjustment during the sitting-standing conversion, how the AI provides timely support for the waist, knees, spine and other parts by predicting the sitting-standing conversion actions. Demonstrates the flexible adjustment of the EAP support material during the conversion process, such as the inflation / deflation of the waist and knee support air cushions.

[0152] For example Figure 7 , the AI model achieves precise adaptation to the user's posture through continuous learning and optimization. During use, the model will be adaptively adjusted in the following ways:

[0153] Personalized learning: The AI system will make personalized adjustments according to the user's body type, habits and health conditions. For example, for users with specific diseases (such as arthritis), the AI model can adjust the support strength by learning to relieve discomfort.

[0154] Real-time data feedback: The AI system will continuously collect data from cameras, pressure sensors and motion sensors, and make real-time adjustments to the support system through optimized algorithms to ensure that each posture adjustment can meet the user's needs.

[0155] Figure 7 Shows how the AI model adjusts the support parameters by real-time monitoring and learning of user data. In the form of a feedback loop, the AI optimizes the support adjustment according to the changes of different users. Demonstrates how to adjust the intensity of each support point according to personalized needs to ensure the personalization and comfort of the user experience.

[0156] For example Figure 8 , the working process of the AI model Figure 8 Shows the working process of how the AI model adjusts various support parameters in real time during sitting, standing and sitting-standing conversion, including data collection, processing, prediction and feedback mechanisms.

[0157] The embodiments described above are only descriptions of the preferred embodiments of the present invention and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations, modifications, and substitutions made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A dynamic support type adaptive scooter based on artificial intelligence, characterized in that, It includes a vehicle body, a 360° somatosensory camera system, an intelligent AI model, and an electroactive polymer support system; The 360° somatosensory camera system is arranged on the vehicle body and is used to capture the user's posture; The electroactive polymer support system is arranged on the vehicle body; The intelligent AI model is connected to the 360° somatosensory camera system and the electroactive polymer support system; the intelligent AI model is used to adjust the electroactive polymer support system according to the user's posture to adapt to different postures of the user.

2. The dynamic support type adaptive scooter based on artificial intelligence according to claim 1, characterized in that, The 360° somatosensory camera system includes two cameras, which are respectively arranged at the upper and lower positions of the vehicle body and are used to capture the upper body posture and lower body posture of the user.

3. The dynamic support type adaptive scooter based on artificial intelligence according to claim 1, characterized in that The vehicle body includes a seat, a front baffle, and a footboard; the front end of the footboard is fixedly connected to the bottom of the front baffle, and the rear end of the footboard is fixedly connected to the seat; a rear wheel is arranged below the seat, and a front wheel is arranged below the footboard.

4. The dynamic support type adaptive scooter based on artificial intelligence according to claim 1, characterized in that, The electroactive polymer support system includes a chest support, leg pads, foot pads, and a rear back pad installed on the vehicle body; the chest support and the rear back pad are arranged opposite to each other and are spaced apart.

5. The dynamic support type adaptive scooter based on artificial intelligence according to claim 1, characterized in that The intelligent AI model includes a vision transformer, a graph neural network, and a long short-term memory network.

6. The dynamic support type adaptive scooter based on artificial intelligence according to claim 5, characterized in that, The input of the vision transformer is the user image data of the 360° somatosensory camera system, and it is used to extract global features from the image and capture the posture changes of the user; The vision transformer divides each frame of image into image patches of a fixed size through image patching, each patch of image is converted into a vector through linear projection, and position encoding is added to retain the spatial position information of the image patches; The vision transformer uses the self-attention mechanism for multi-layer feature learning to capture the global semantic information in the image; The vision transformer outputs high-level features of the image extracted by multiple self-attention layers.

7. The dynamic support type adaptive scooter based on artificial intelligence according to claim 5, characterized in that, The input of the graph neural network is the user image data of the 360° somatosensory camera system, and it is used to capture the spatial relationships and dependencies between human joints; The graph neural network initializes each human joint as a node in the graph through node initialization, and the features of the node include the coordinates, angles, and speed information of the joint; The graph neural network propagates information through graph convolutional layers to calculate the spatial dependency relationships between each joint; The graph neural network generates spatial dependency features between joints and transmits them to the long short-term memory network for temporal modeling.

8. The dynamic support type adaptive walking aid based on artificial intelligence according to claim 5, characterized in that, The input of the long short-term memory network comes from the image features and spatio-temporal features of the vision transformer and the graph neural network; it is used to model the movement and posture changes of the human body in the time dimension and capture long-term dependencies; The long short-term memory network receives the spatio-temporal features from the vision transformer and the graph neural network and processes time series data; The long short-term memory network captures time series dependencies through an internal gating mechanism; The long short-term memory network generates posture categories and support strengths and feeds them back to the electroactive polymer support system to adjust the electroactive polymer support system.

Images